SatDepth: A Novel Dataset for Satellite Image Matching

2025-03-17

AI Summary

This work addresses the challenge of pixel-level correspondence estimation in satellite image matching under large viewpoint changes, multi-temporal acquisitions, and strong rotational disparities. To this end, we introduce SatDepth—the first dense ground-truth dataset specifically designed for satellite imagery. Methodologically, we propose a rotation-augmented balanced sampling strategy, enabling robust training and evaluation under substantial rotational differences for the first time. Dense disparity ground truth is generated via multi-revisit image registration. We conduct systematic benchmarking and ablation studies on state-of-the-art deep matching models, including SuperPoint and LoFTR. Experimental results demonstrate that rotation-augmented training improves matching accuracy by up to 40%, significantly outperforming models trained on terrestrial image datasets. SatDepth establishes a new benchmark and provides a practical technical pathway for satellite image matching.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Large-scale security measurements
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent advances in deep-learning based methods for image matching have demonstrated their superiority over traditional algorithms, enabling correspondence estimation in challenging scenes with significant differences in viewing angles, illumination and weather conditions. However, the existing datasets, learning frameworks, and evaluation metrics for the deep-learning based methods are limited to ground-based images recorded with pinhole cameras and have not been explored for satellite images. In this paper, we present ``SatDepth'', a novel dataset that provides dense ground-truth correspondences for training image matching frameworks meant specifically for satellite images. Satellites capture images from various viewing angles and tracks through multiple revisits over a region. To manage this variability, we propose a dataset balancing strategy through a novel image rotation augmentation procedure. This procedure allows for the discovery of corresponding pixels even in the presence of large rotational differences between the images. We benchmark four existing image matching frameworks using our dataset and carry out an ablation study that confirms that the models trained with our dataset with rotation augmentation outperform (up to 40% increase in precision) the models trained with other datasets, especially when there exist large rotational differences between the images.

Problem

• Lack of datasets for satellite image matching
• Challenges in handling rotational differences in satellite images
• Need for improved deep-learning frameworks for satellite imagery

Innovation

• Introduces SatDepth dataset for satellite image matching
• Proposes rotation augmentation for handling image variability
• Benchmarks frameworks, shows 40% precision improvement